SPINE is a repo-native mission governor for AI coding workflows. It sits above coding agents and helps keep work:
- focused
- bounded
- reviewable
- evidence-backed
- resistant to drift
This roadmap explains:
- where SPINE is today
- what each phase means and when it is complete
- what "v1.0.0" should actually mean
SPINE has exited beta. v0.2.0 is the current stable release (2026-04-10).
Alpha exit is complete. Phase 3A delivered:
- explicit repo targeting +
--cwdprecedence normalization - repo/branch context visibility + deterministic default branch resolution
- stable exit codes +
--jsonoutput modes - bootstrap polish + discipline-tax ergonomics
- artifact ergonomics contract
- external-repo onboarding docs
The current stage is:
**Post-Beta — v0.2.0 stabilization. Bug fixes only. Phase 3B candidates deferred.**d for agents and operators without hiding what happens.
SPINE is not trying to become another coding agent, a dashboard product, or a cloud control plane.
Its direction is:
- governance above generation
- repo-native truth
- deterministic, inspectable behavior
- clear human authority
- agent-friendly execution surfaces
A core product rule guides this roadmap:
Agents may execute governance mechanics. Operators retain governance authority.
That means SPINE should reduce operational friction without hiding what happened, silently mutating repo truth, or replacing operator judgment.
Complete. Alpha exit criteria were met and v0.2.0-beta was released on 2026-04-07.
All Phase 3A goals delivered:
- repo targeting is explicit and predictable
- repo / branch / compare-context visibility is trustworthy
- machine-readable output exists where justified
- exit behavior is stable and documented
- artifact references are easy to inspect and automate against
- bootstrap/install guidance is clear for non-SPINE repos
- external-repo docs/examples are strong
- no unwanted scope drift crept in
Complete (2026-04-09). Beta exit validation passed. Beta exit criteria met:
- 0 known blocker bugs on shipped core surface
- 0 high-severity trust/usability bugs on core agent workflows
- Docs, command examples, and CLI tell the same truth
- SPINE's own repo is self-governed with real mission, evidence, decisions
- Repeated governed workflows validated end-to-end (29-step harness, 2 iterations)
v0.2.0 released 2026-04-10. Beta work focused on reducing discipline tax and making governance mechanics more executable for agents and tools.sting ceremony**.
This is where SPINE should more seriously reduce discipline tax.
Discipline tax is the accumulated effort cost of using governance consistently over time.
If governance requires too much manual repetition, people will:
- agree with it
- admire it
- then skip parts of it in real work
Beta should attack that problem directly.
SPINE should reduce discipline tax not by hiding governance, but by making governance easier for agents and tools to execute explicitly.
That means "agent-friendly" should look like:
- stable command contracts
- machine-readable outputs
- deterministic repo targeting
- visible context
- predictable artifact locations
- auditable state changes
- easy handoff between tools
It should not look like:
- hidden background behavior
- silent auto-logging
- invisible scope changes
- automation that obscures authority or truth
Make it easier to govern bounded work without polluting canonical mission state.
Possible outcomes:
- issue-scoped brief generation
- task-scoped governance lanes
- clearer branch-local workflows
Reduce discipline tax where real work changes hands.
Possible outcomes:
- PR governance summaries
- evidence rollups
- handoff summaries
- "what changed against mission" reports
Turn governance into explicit checkpoints.
Possible outcomes:
- before-work checks
- before-PR checks
- after-test checks
Let agents draft governance records without becoming the final authority.
Possible outcomes:
- draft evidence helpers
- draft decision helpers
- clearer proposed-vs-canonical distinction
Improve local agent/tool consumption of SPINE state without drifting into cloud/platform territory.
Possible outcomes:
- richer local MCP resources
- stable schemas for current mission, latest brief, open drift, latest review, recent evidence/decisions
Make branch-local governance safer and less error-prone.
Possible outcomes:
- keep-local vs promote-canonical flows
- protections against accidental canonical churn
Beta is working if:
- agents can execute governance mechanics reliably
- humans still control mission and scope authority
- repeat use feels lighter, not heavier
- governance remains repo-native and inspectable
- SPINE is clearly useful on repos beyond SPINE itself
v1.0.0 does not mean "finished forever."
It means:
- the core governance model is stable
- the trust model is stable
- the file/state contracts are stable
- the tool is dependable enough for repeated real use
The core CLI should feel settled and dependable.
The .spine/ model should be documented, durable, and migratable.
The line between:
- agent-executable governance mechanics
- operator-authorized governance decisions
must be clear and trustworthy.
Machine-readable outputs, exit codes, artifact conventions, and local tool-consumption surfaces should be stable enough for real tooling to rely on.
SPINE should reach 1.0 when there is evidence of repeated use:
- on SPINE itself
- on unrelated repos
- across real AI-assisted development workflows
By v1, docs should cleanly cover:
- quickstart
- external-repo usage
- command reference
- governance model
- architecture/spec
- troubleshooting
- migration/versioning
- advanced workflows
By v1, the repo and release process should feel normal, boring, and dependable:
- clean milestones
- versioned releases
- issue hygiene
- release-note discipline
- contribution/security/community docs
- solid CI
SPINE is ready for 1.0.0 when:
- the governance model is stable
- the authority boundaries are stable
- the repo-native trust model is stable
- the automation-facing surfaces are stable
- repeated use on real repos is demonstrated
- the open-source maintenance story is credible
Minimum credible baseline:
- public repo
- clean default branch
- branch protection
- release tags
LICENSESECURITY.md- CI
- roadmap/backlog/status docs
- milestone and issue tracking
Expected improvements:
- issue templates
- PR template
- contribution guide
- changelog discipline
- stronger CI coverage
- better docs index and contributor guidance
- clearer maintenance expectations
Expected maturity:
- dependable release cadence
- documented migration/versioning expectations
- durable issue/milestone hygiene
- strong community and security surfaces
- repo/org state that looks normal and trustworthy
SPINE should reduce discipline tax by making governance easier to execute explicitly, especially for agents and tools, while keeping authority, truth, and auditability in the repo.
- less repeated manual ceremony
- less ambiguity about operating context
- less repeated re-explanation
- easier-to-find governance artifacts
- easier handoff and PR preparation
- clearer CI usage
- hidden state
- silent canonical mutation
- fake automation theater
- invisible scope changes
- replacing human judgment with opaque behavior
In Alpha — reduce obvious friction:
- targeting
- context visibility
- structured output
- bootstrap clarity
- examples
- artifact discoverability
In Beta — make governance mechanics more executable for agents:
- task lanes
- handoff/PR-prep
- draftable records
- preflight/checkpoint flows
- stronger local tool surfaces
In v1 — stabilize these behaviors so they are predictable, inspectable, and dependable.
These remain out of scope unless a future spec explicitly changes direction:
- dashboard-first control surfaces
- cloud sync / cloud control plane
- remote MCP hosting
- account systems
- billing
- multi-user collaboration
- autonomous multi-agent orchestration
- hidden background governance daemons
- model-required core governance logic
- silent canonical state mutation
- broad platform sprawl disguised as "ergonomics"
SPINE should reduce friction without reducing visibility.
- Define the first bounded beta implementation slice(s)
- Open the Beta milestone on GitHub with scoped, workable issues
- Begin Beta execution — discipline-tax reduction, agent-executable mechanics, handoff primitives
SPINE is not trying to become another coding agent or an orchestration platform.
It is trying to become the trustworthy governance layer above AI coding tools:
- repo-native
- explicit
- inspectable
- auditable
- sustainable in repeated real use